- Email: [email protected]
Visual field differences in the processing of numerical stimuli
BRAIN
AND
COGNITION
7, 247-256 (1988)
Visual Field Differences in the Processing of Numerical Stimuli RAYMOND KLEIN AND JULIA MCINNES Dalhousie
University,
Halifax,
Nova Scotia, Canada
Twenty-four right-handed subjects received random presentations of the numbers l-6 in the form of words, digits, and dot patterns, to the left and right visual fields. Accuracy and reaction time were recorded for an odd-even judgment requiring a manual response. A significant stimulus type of visual field interaction was obtained, with words showing a left-hemisphere advantage and digits and dot patterns showing a right-hemisphere advantage. This pattern supports Coltheart’s (1980, Deep dyslexia: A right hemisphere hypothesis, In M. Coltheart, K. Patterson, & J. C. Marshall (Eds.), Deep dyslexia, London: Routledge & Kegan Paul) right hemisphere reading hypothesis, which suggests that the left hemisphere’s general advantage in processing linguistic material may be specific to stimuli which involve phonological processing. When phonological processing is not possible (e.g., for arabic digits and other ideographic orthographies), the right hemisphere may have an advantage because of its superior visuospatial processing capabilities. 0 1988 Academic Press, Inc.
In terms of visual processing, traditional views of hemispheric asymmetry in the human brain have characterized the left hemisphere as being specialized for processing written language while the right hemisphere has been linked to superior visuospatial processing capabilities (Bryden, 1982). In fact, it has been proposed that the right hemisphere suffers from “word blindness” (Geschwind, 1965), which refers to an inability to process written language. This latter point has been challenged by Coltheart (1980), who has developed a right hemisphere reading hypothesis to help account for the syndrome of deep dyslexia. Coltheart (1980) suggests that although the right hemisphere may be incapable of phonological recoding, which is considered to be an important strategy for normal reading, it may be capable of processing written language using This paper is based on an undergraduate honors thesis by J. McInnes (1985) under the supervision of R. Klein. The two authors contributed equally to the paper in its present form. Requests for reprints should be addressed to R. Klein, Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada B3H 451. 247 0278-2626/88$3.00 Copyright 0 1988 by Academic Press, Inc. All rights of reproduction in any form reserved.
248
KLEIN
AND
MC INNES
a visuospatial analysis of words. Since the reading deficits observed in deep dyslexics typically involve phonological recoding skills, Coltheart (1980) suggests that these individuals may be reading with their right hemisphere. Some support for this hypothesis has come from clinical studies of brain-damaged patients (but see Patterson & Besner, 1984, for a different interpretation). Smith (1966) describes an individual who, after surgical removal of the left hemisphere, was able to comprehend color words and some object names, Levy and Trevarthen (1977) reported that, although their patients could comprehend words which were directed to the right hemisphere, they could not judge whether the words rhymed with a previously spoken word. The left hemisphere was able to perform both tasks. These studies suggest that the right hemisphere may possess mechanisms which allow for a direct route from print to meaning, in the absence of an ability to appreciate the phonological characteristics of words. Further evidence in support of Coltheart’s (1980) right hemisphere reading hypothesis has come from laterality studies involving two Japanese orthographies, or writing systems: Kanji and Kana. Kanji is an ideographic system in which words are represented as individual characters and so pronunciation may be considered to be arbitrarily assigned to each character. Kana, by contrast, is a syllabic system in which each character corresponds to a particular syllable. In terms of visual processing of these two orthographies, Kana may make use of phonological recoding while Kanji must rely on a direct linkage between visuospatial information and meaning. Some laterality studies have demonstrated a left-hemisphere advantage in the recognition of Kana stimuli and a right-hemisphere advantage in the processing of Kanji stimuli (Hatta, 1977; Sasanuma, Itoh, Mot-i, and Kobayashi, 1977), indicating that, in the absence of phonological information, the right hemisphere may be superior in analyzing written language. In the English language, the arabic digits represent an ideographic system which is similar to Kanji characters in their lack of phonological information. This similarity has led researchers to suggest that digits may show evidence of right-hemisphere processing (Coltheart, 1980). Since numbers in English may be represented alphabetically or ideographically (e.g., one, l), they permit an evaluation of Coltheart’s (1980) right hemisphere reading hypothesis, which avoids the problems of the generalizability of data from brain-damaged and non-English speaking individuals. Studies which have investigated the asymmetries involved in digit processing have, up to now, been scarce and the results have been inconclusive. However, in light of Coltheart’s (1980) hypothesis, it is of interest to examine the laterality literature involving arabic digits, since it allows for an evaluation of the asymmetries in processing ideographic stimuli.
PROCESSING
NUMERICAL
STIMULI
249
Diamond and Beaumont (1972) carried out one of the earliest studies exploring lateralization of digit processing using a mental arithmetic task. They presented subjects with three double-digit numbers to the left or right visual fields. The subjects were instructed to add or subtract the first two numbers and to compare the result with the third number. A right-hemisphere advantage (in terms of accuracy) was reported for trials involving subtraction, while the visual field difference for addition did not reach statistical significance. The researchers argued that the relative ease of performing the addition task resulted in a failure to show any visual field effects. Since an exposure duration of 640 msec was used in the study, the possibility of saccadic eye movements cannot be ruled out. A similar paradigm was employed by Hoff and McKeever (1979). They presented three single-digit numbers for 100 msec each and reported a left-hemisphere advantage for addition problems. However, their use of a verbal response may have introduced a bias towards left-hemisphere processing. Troup, Bradshaw, and Nettleton (1982) attempted to control for this problem by using a bimanual response. This investigation resulted in a right-hemisphere advantage (based on accuracy) for addition and multiplication problems. These studies indicate that the right hemisphere may play a role in the processing of digit stimuli, when mental arithmetic is involved. It is unclear, however, whether the right-hemisphere involvement is due to the ideographic nature of the digits or to the mechanisms involved in calculation. Two studies have employed a comparative judgment task, which involves deciding which of two digits is numerically larger. Since this task can be considered to be independent of calculation, it may provide a more accurate measure of the differential stimulus processing of digits by the two hemispheres. Besner, Grimsell, and Davis (1979) reported a lefthemisphere advantage for this task, while Katz (1980) observed a righthemisphere advantage. Katz (1980) suggested that methodological differences involving stimulus size, exposure duration and inter-trial intervals may have been responsible for the discrepancies in the findings of the two studies. Discrimination and identification tasks represent a further simplification of tasks that may be used to assess the asymmetries of digit processing. Besner, Daniels, and Slade (1982) employed random unilateral presentations of single digits and asked subjects to identify each digit using a written response. They reported a strong left-hemisphere advantage, which may have been produced by a left-hemisphere bias due to oral mediation of the written response and/or the use of only right-hand responding. Cohen’s (1975) investigation of numerical stimuli permits the most direct evaluation of Coltheart’s (1980) hypothesis since it involves a comparison of alphabetic, ideographic, and pictographic representations of the same numerical stimuli. She asked subjects to discriminate, using
250
KLEIN
AND
MC INNES
a bimanual response, between the stimuli 4 and 5, which were presented as words, digits, or dice patterns. When these stimuli were randomly presented without precues, no visual field differences were observed for any of the stimulus types. However, when subjects were provided with a cue indicating what the next stimulus type would be, a left-hemisphere advantage for words and digits (which were preceded by verbal cues) and a right-hemisphere advantage for dice (nonverbal cue) emerged. While the finding that the digit stimuli produced a left-hemisphere advantage would seem to be strong evidence against Coltheart’s (1980) hypothesis, these stimuli were cued with the word “DIGIT,” which may have introduced a left-hemisphere bias for that stimulus type (e.g., see Kershner, Thomas, & Calloway , 1977). Additionally, the use of a very limited stimulus set (only two numbers) might have encouraged a strategy of looking for distinctive features (such as the presence vs. absence of a curved line, which would be diagnostic for discriminating 4 and 5) that might favor the left hemisphere. To summarize, the results of studies that have investigated the asymmetries involved in digit processing have, up to now, been inconclusive. The methodological issues of task requirements and response selection may contribute significantly to these inconsistencies. Hemispheric biasing effects may be reduced by employing a simple task using a manual response (counterbalanced for hand) and random presentations of stimulus types to ensure that stimulus type differences are real and not due to attentional biases (see Kinsbourne, 1970). The present study was conducted to overcome some of the methodological problems associated with earlier studies of hemispheric differences in digit processing. The stimulus types (words, digits, dot patterns used on dice) employed in the Cohen study (1975) were chosen because they allow for a comparison of alphabetic, ideographic, and pictographic representations of the same numerical stimuli, thus permitting a direct evaluation of Coltheart’s (1980) right hemisphere reading hypothesis. To rule out differential attentional biases across these stimulus types, they were randomly mixed. A simple odd-even judgement task was chosen for the current study because it allows for a brief, single stimulus presentation which requires identification without the complexity of calculation. Additionally, the task permits a two-choice, unimanual response, which avoids the hemispheric biasing effects produced by a verbal response. The stimulus set used in the Cohen study (1975) was expanded to include the numbers 1 through 6, so that the use of any strategies that may have been induced by a limited stimulus set (e.g., feature extraction) could be ruled out. Based on Coltheart’s (1980) right hemisphere reading hypothesis, it was predicted that words would show a left-hemisphere advantage because they permit phonological processing. Since phonological processing is not possible for the digit and dice stimuli, it was predicted
PROCESSING NUMERICAL
FIG. I.
STIMULI
251
Examples of dot/dice, digit, and word stimuli used in the experiment.
that they would show a right-hemisphere advantage, due to the right hemisphere’s superior visuospatial processing capabilities. METHOD Subjects. Twenty-four undergraduate students (12 females; 12 males), all of whom were native English speakers, participated in the study for course credit. A handedness inventory demonstrated that all subjects were strongly right-handed. All had normal or correctedto-normal vision. Apparatus. The stimuli were displayed on a Tektronix 604 display monitor with a P31 phosphor, under the control of a PDP-II:10 computer. Subjects viewed the display from a distance of 2.5 ft. The stimuli were the numbers 1-6, displayed in the form of vertically oriented words, arabic digits, and dot patterns (those used on dice). The stimuli are shown in Fig. 1. The words were written in uppercase dot matrix letters each of which subtended about 0.17 deg horizontally and 0.2 deg vertically. The arabic numerals subtended 0.34 deg horizontally and 0.4 deg vertically. Each dot in the dice stimuli subtended about 0.08 deg, and the full matrix subtended 0.36 deg vertically and horizontally. Procedure. Each subject was required to judge whether a stimulus flashed to the right or left of a central fixation point was odd or even. The response involved pressing one of two keys with either the middle or index finger of one hand. Reaction time and accuracy were recorded for each trial. A trial proceeded as follows: a central fixation point was displayed for 2 sec. The fixation point then increased in brightness for 0.5 set, as a warning to subjects that the stimulus was about to appear. The stimulus was then presented to either the right or left of fixation for 160 msec after which the screen became blank. Subjects were allowed up to 3 set in which to make their response. Following the response, there was a 2-set intertrial interval before the next trial began. The order of presentation of stimulus type and the visual field of presentation were randomized. Each stimulus type occurred equally often in the left and right visual field.
252
KLEIN AND MC INNES -.,
I &II /
IS.91 COTS
15%
(4.9) WORos
LVF
RVF
800 2 I
t
FIG. 2. Performance as a function of visual field and stimulus type. Percentage error for each condition is shown in parentheses. Each subject received 36 practice trials following general instructions. The instructions emphasized that the subject should maintain her/his focus on the central fixation point and respond as quickly and accurately as possible to the stimulus. Following the practice trials, there were two blocks of 144 experimental trials each (with a 5-min rest period between blocks). In each block of 144 trials, each stimulus type occurred 24 times in each visual field. Each number occurred equally often in each combination of visual field and stimulus type. Subjects were required to keep their two fingers resting on the keys throughout the experiment. The response hand was counterbalanced between subjects such that half responded with their right hand and the other half with their left hand. For each of these two groups, half used their index finger for odd responses and their middle finger for even responses. The other half used the opposite arrangement.
RESULTS Mean correct reaction time (RT) and mean accuracy (% correct) were each subjected to a five-way analysis of variance (sex x hand x block x stimulus type x visual field). Analysis of the reaction time data (each entry in the ANOVA was based on up to 36 trials, depending on the number of errors) revealed main effects of block (F( 1, 20) = 13.38, p < .Ol) and stimulus type (F(2, 40) = 105.75, p < .OOOl).There were no main effects of sex, hand, or visual field. The only significant interactions involved stimulus type and visual field (F(2, 40) = 3.27, p < .OS) and sex and stimulus type (F(2, 40) = 6.11, p < .Ol). The stimulus type x visual field interaction is shown in Fig. 2. These data suggest that there is a tendency for the right visual field (RVF) to be faster with words and the left visual field (LVF) to be faster with digits and dots, exactly the pattern of results predicted by Coltheart’s hypothesis. Planned contrasts revealed that the visual field differences for each type of stimulus were not significant (words, F(1, 40) = 2.06; numerals, F(1, 40) = 3.69; dots, F(1, 40) = 1.98), but the visual field effect for the word stimuli was significantly different from that for the numerals and dots combined (F( 1,
PROCESSING
NUMERICAL
STIMULI
253
40) = 6.41, p < .025). When contrasted separately, the difference between words and numerals was significant (F(1, 40) = 5.63, p < .05) while that between words and dots was nearly significant (F(1, 40) = 4.04, p = .0512). The effect of block revealed that subjects were faster in the second block. The interaction between sex and stimulus type was due to the fact that males processed dot patterns and digits with equal speed while females were slower on the dot patterns. The analysis of accuracy revealed significant main effects of sex (F( 1, 20) = 4.78, p < .05), block (F(1, 20) = 16.13, p < .OOl>and stimulus type (F(2,40) = 14.96, p < ,001). No interactions were significant. These effects indicate that males were more accurate than females, performance accuracy improved with block, and accuracy was best for digits, worst for dot patterns, and intermediate for words. Although the interaction between visual field and stimulus type was not significant (F(1) 40) = l&3), these data are shown parenthetically in Fig. 1 to permit evaluation of the possibility that the obtained interaction in RT might be due to speed-accuracy tradeoffs. As can be seen, in each visual field comparison, the faster RTs are associated with more accurate performance, so the RT pattern is clearly nor compromised by accuracy tradeoffs. Within each type of stimulus, RTs for each numerical quantity (l-6) were combined across the two blocks and subjected to ANOVAS with visual field and quantity as variables (each entry in the ANOVA was based on up to 12 trials). Significant performance variations as a function of quantity were observed for each stimulus type: With words (F(5, 115) = 3.1, p < .025), the ordering of RTs was 6 < 1, 2, 3, 4, < 5; with numerals (F(5, 115) = 5.65, p < .Ol) the ordering was 2, 4, 6 < 1, 3, 5; and with dots (F(5, 115) = 2.99, p < .025) the ordering was 1, 4 < 2, 3, 6 < 5. However, in no case did this pattern interact with visual field (all F’s < 1). DISCUSSION
The interaction between stimulus type and visual field observed in reaction time offers support for Coltheart’s (1980) right hemisphere reading hypothesis. The word stimuli, which contain explicit phonological information, produced a RVF advantage of 12 msec. Conversely a LVF advantage of comparable magnitude was observed for the digit and dot stimuli, both of which may be characterized as lacking phonological information. These findings are open to several interpretations. If one assumes that the right hemisphere is incapable of processing linguistic information, due to its inability to perform phonological recoding, it follows that words which are presented to the LVF must be transferred to the left hemisphere in order to be processed. In this case, the 12msec visual field difference observed for word stimuli may reflect the
254
KLEIN AND MC INNES
time to transfer the graphological information from the right to the left hemisphere for phonological recoding. The corresponding decrease in accuracy (- 1.3%) may reflect the loss of information during transfer. Alternatively, the right hemisphere may utilize a strategy for identifying the words in order to make the odd/even judgment which does not require phonological recoding but which takes longer than the left hemisphere’s processing strategy. The tendency for digits and dots to be processed faster by the right hemisphere may be interpreted in a similar fashion. The observed 16msec visual field difference may reflect a transfer time of visual information from the left to right hemisphere, or it may reflect a slower and possibly different nonphonological processing strategy by the left hemisphere. These possibilities cannot be distinguished using the present data, but the fact that RT differences as a function of numerical quantity did not interact with visual field seems more consistent with an explanation based on transfer. A recently published study by Besner, Snow, and Davelaar (1986) merits discussion since it addressed the question of differential hemispheric processing of alphabetic and ideographic orthographies using a different methodology and reports different results from the present study. They compared performance on unilaterally presented single letters and digits using a verbal identification response. The major rationale seems to have been that clinical studies of split-brain patients have employed similar comparisons yielding right-hemisphere advantages for letters, but that this finding cannot be generalized to the normal population. Indeed, unlike the expectation based on the clinical literature, Besner et al. (1986) found equivalent RVF advantages for the two types of material. The authors propose a semantic distinction between single letters and numerals, with the former being void of semantic information and they argue that it is this factor which is responsible for the differential hemispheric processing in the clinical literature. While it is true that digits contain more concrete semantic information, it may be argued that single letters contain semantic associations in the form of prototype words (e.g., A = apple, C = cat). In any case, to the extent that the clinical literature has compared single letters and digits in an identification task, Besner et al. (1986) have provided a useful extension to the normal population. However, as a test of Coltheart’s hypothesis the Besner et al. (1986) study is flawed. In the first place, the appropriateness of using single letters to represent an alphabetic script must be questioned since the arbitrary assignment of pronunciation to each letter characterizes them as ideographic. A direct comparison of ideographic and alphabetic scripts is better achieved by using digits and their alphabetic counterparts (i.e., 1, one) as in the present experiment. Second, there is the very real possibility that their observed left-hemisphere advantage for letters and
PROCESSING
NUMERICAL
STIMULI
255
numerals is due to a biasing effect inherent in the use of a vocal identification response. At the very least, such a bias could override any visual field differences that may exist for the two stimulus types (see, e.g., Hellige & Sergent, 1986), and for this reason their failure to obtain a visual field by type of stimulus interaction is not particularly informative. Given our choice of materials, there is the inevitable fact that the stimuli are physically different. While the stimuli were roughly equated for visibility and were composed of the same size luminous points, the density of these points and the overall size of the stimuli were somewhat different for the word, number, and dot stimuli. In light of the proposal that the two hemispheres might differ in their ability to process high and low spatial frequencies (Sergent & Hellige, 1986) it might be suggested that the pattern we have observed is due to the relative weight of high and low spatial frequencies in the composition of our stimulus materials. We do not favor this interpretation for two reasons. First, although direct manipulation of the size of stimuli has produced changes in hemispheric asymmetries (Sergent, 1983; Pring, 1981), the spatial frequency hypothesis has not received direct support from research on the detection of sine wave gratings (Kitterle & Kaye, 1985). Second, an analysis of the spatial frequency composition of our stimuli reveals a great degree of overlap, and thus does not permit a straightforward explanation of our data in terms of the spatial frequency hypothesis. Nevertheless, although we prefer an explanation based on Coltheart’s hypothesis, we cannot rule out the possibility that physical differences between our stimuli might be partly responsible for the observed visual field by type of stimulus interaction. The role of the odd/even judgement task in mediating the observed interaction merits some discussion. It is possible that one, or the other, hemisphere is superior at performing the odd/even judgement. This could have the effect of biasing performance toward the superior hemisphere. For example, the fact that the left hemisphere was not significantly better than the right with the word stimuli suggests that the right hemisphere might be superior on this task. From the point of view of the present experiment, however, which does not permit a direct assessment of this possibility, the important observation is that the visual field effects observed depend on the type of representation of the numerical stimuli (i.e., the visual field by type of stimulus interaction). The odd/even task was chosen for the current study because it allows for a manual response, a short exposure duration, presentation of a single item, and is relatively simple. In future work it would be useful to investigate different tasks which meet these criteria, to determine the generality of the observed interaction between visual field and stimulus type in the processing of numerical stimuli.
KLEIN AND MC INNES
256
REFERENCES Besner, D., Daniels, S., & Slade, C. 1982. Ideogram reading and right hemisphere language. British Journal of Psychology, 73, 21-28. Besner, D., Grimsell, D., & Davis, R. 1979. The mind’s eye and the comparative judgement 17, 373-380. of number. Neuropsychologia, Besner, D., Snow, D., & Davelaar, E. 1986. Logographic reading: Is the right hemisphere special? Canadian Journal of Psychology, 40, 45-53. Bryden, M. P. 1982. Functional asymmetry in the intact brain. New York: Academic Press. Cohen, G. 1975. Hemisphere differences in the effects of cuing in visual recognition tasks. Journal
of Experimental
Psychology:
Human
Perception
and Performance,
1, 366-
373. Coltheart, M. 1980. Deep dyslexia: A right hemisphere hypothesis. In M. Coltheart, K. Patterson, & J. C. Marshall (Eds.), Deep dyslexia. London: Routledge & Kegan Paul. Dimond, S., & Beaumont, J. 1972. A right hemisphere basis for calculation in the human brain. Psychonomic Science, 26, 137-138. Geschwind, N. 1965. Disconnection syndromes in animals and man. Brain, 88, 327-394 and 585-644. Hatta, T. 1977.Recognition of Japanese Kanji in left and right visual field. Neuropsychologia, 15, 685-688.
Hellige, J., & Sergent, J. 1986. Role of task factors in visual-field asymmetries. Brain and Cognition,
5, 200-223.
Hoff, A. L., & McKeever, W. F. 1979. Hemispheric specialization for simple arithmetic calculation: evidence from normal subjects and from a cerebral commissurotomy patient. Clinical
Neuropsychology,
I, 24-28.
Katz, A. 1980.Cognitive arithmetic: Evidence for right hemisphere mediation in an elementary component stage. Quarterly Journal of Experimental Psychology, 32, 69-84. Kershner, J., Thomas, R., & Callaway, R. 1977. Nonverbal fixation control in young children induces a left-field advantage in digital recall. Neuropsychologia, 15, 569576. Kinsbourne, M. 1970. The cerebral basis of lateral asymmetries in attention. In A. F. Sanders (Ed.), Attention and performance III. Amsterdam: North-Holland. Kitterle, F. L., & Kaye, R. S. 1985. Hemispheric symmetry in contrast and orientation sensitivity. Perception & Psychophysics, 37, 391-396. Levy, J., & Trevarthen, C. 1977. Perceptual, semantic and phonetic aspects of elementary language processes in split brain patients. Brain, 100, 105-118. Patterson, K., & Besner, D. 1984.Is the right hemisphere literate? Cognitive Neuropsychology, 1, 315-341. Pring, T. R. 1981. The effects of stimulus size and exposure duration on visual field asymmetries. Cortex, 17, 227-240. Sasanuma, S., Itoh, M., Mori, K., & Kobayashi, Y. 1977. Tachistoscopic recognition of Kana and Kanji words. Neuropsychologia, 15, 547-553. Sergent, J. 1983. The effects of sensory limitations on hemispheric asymmetries. Canadian Journal
of Psychology,
37, 345-366.
Sergent, J., & Hellige, J. 1986. Role of input factors in visual-field asymmetries. Brain and Cognition, 5, 174-199. Smith, A. 1966. Speech and other functions after left (dominant) hemispherectomy. Journal of Neurology, Neurosurgery and Psychiatry, 29, 467-471. Troup, G. A., Bradshaw, J. L., & Nettleton, N. C. 1983. The lateralization of arithmetic and number processing: A review. International Journal of Neuroscience, 19, 321242.
AND
COGNITION
7, 247-256 (1988)
Visual Field Differences in the Processing of Numerical Stimuli RAYMOND KLEIN AND JULIA MCINNES Dalhousie
University,
Halifax,
Nova Scotia, Canada
Twenty-four right-handed subjects received random presentations of the numbers l-6 in the form of words, digits, and dot patterns, to the left and right visual fields. Accuracy and reaction time were recorded for an odd-even judgment requiring a manual response. A significant stimulus type of visual field interaction was obtained, with words showing a left-hemisphere advantage and digits and dot patterns showing a right-hemisphere advantage. This pattern supports Coltheart’s (1980, Deep dyslexia: A right hemisphere hypothesis, In M. Coltheart, K. Patterson, & J. C. Marshall (Eds.), Deep dyslexia, London: Routledge & Kegan Paul) right hemisphere reading hypothesis, which suggests that the left hemisphere’s general advantage in processing linguistic material may be specific to stimuli which involve phonological processing. When phonological processing is not possible (e.g., for arabic digits and other ideographic orthographies), the right hemisphere may have an advantage because of its superior visuospatial processing capabilities. 0 1988 Academic Press, Inc.
In terms of visual processing, traditional views of hemispheric asymmetry in the human brain have characterized the left hemisphere as being specialized for processing written language while the right hemisphere has been linked to superior visuospatial processing capabilities (Bryden, 1982). In fact, it has been proposed that the right hemisphere suffers from “word blindness” (Geschwind, 1965), which refers to an inability to process written language. This latter point has been challenged by Coltheart (1980), who has developed a right hemisphere reading hypothesis to help account for the syndrome of deep dyslexia. Coltheart (1980) suggests that although the right hemisphere may be incapable of phonological recoding, which is considered to be an important strategy for normal reading, it may be capable of processing written language using This paper is based on an undergraduate honors thesis by J. McInnes (1985) under the supervision of R. Klein. The two authors contributed equally to the paper in its present form. Requests for reprints should be addressed to R. Klein, Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada B3H 451. 247 0278-2626/88$3.00 Copyright 0 1988 by Academic Press, Inc. All rights of reproduction in any form reserved.
248
KLEIN
AND
MC INNES
a visuospatial analysis of words. Since the reading deficits observed in deep dyslexics typically involve phonological recoding skills, Coltheart (1980) suggests that these individuals may be reading with their right hemisphere. Some support for this hypothesis has come from clinical studies of brain-damaged patients (but see Patterson & Besner, 1984, for a different interpretation). Smith (1966) describes an individual who, after surgical removal of the left hemisphere, was able to comprehend color words and some object names, Levy and Trevarthen (1977) reported that, although their patients could comprehend words which were directed to the right hemisphere, they could not judge whether the words rhymed with a previously spoken word. The left hemisphere was able to perform both tasks. These studies suggest that the right hemisphere may possess mechanisms which allow for a direct route from print to meaning, in the absence of an ability to appreciate the phonological characteristics of words. Further evidence in support of Coltheart’s (1980) right hemisphere reading hypothesis has come from laterality studies involving two Japanese orthographies, or writing systems: Kanji and Kana. Kanji is an ideographic system in which words are represented as individual characters and so pronunciation may be considered to be arbitrarily assigned to each character. Kana, by contrast, is a syllabic system in which each character corresponds to a particular syllable. In terms of visual processing of these two orthographies, Kana may make use of phonological recoding while Kanji must rely on a direct linkage between visuospatial information and meaning. Some laterality studies have demonstrated a left-hemisphere advantage in the recognition of Kana stimuli and a right-hemisphere advantage in the processing of Kanji stimuli (Hatta, 1977; Sasanuma, Itoh, Mot-i, and Kobayashi, 1977), indicating that, in the absence of phonological information, the right hemisphere may be superior in analyzing written language. In the English language, the arabic digits represent an ideographic system which is similar to Kanji characters in their lack of phonological information. This similarity has led researchers to suggest that digits may show evidence of right-hemisphere processing (Coltheart, 1980). Since numbers in English may be represented alphabetically or ideographically (e.g., one, l), they permit an evaluation of Coltheart’s (1980) right hemisphere reading hypothesis, which avoids the problems of the generalizability of data from brain-damaged and non-English speaking individuals. Studies which have investigated the asymmetries involved in digit processing have, up to now, been scarce and the results have been inconclusive. However, in light of Coltheart’s (1980) hypothesis, it is of interest to examine the laterality literature involving arabic digits, since it allows for an evaluation of the asymmetries in processing ideographic stimuli.
PROCESSING
NUMERICAL
STIMULI
249
Diamond and Beaumont (1972) carried out one of the earliest studies exploring lateralization of digit processing using a mental arithmetic task. They presented subjects with three double-digit numbers to the left or right visual fields. The subjects were instructed to add or subtract the first two numbers and to compare the result with the third number. A right-hemisphere advantage (in terms of accuracy) was reported for trials involving subtraction, while the visual field difference for addition did not reach statistical significance. The researchers argued that the relative ease of performing the addition task resulted in a failure to show any visual field effects. Since an exposure duration of 640 msec was used in the study, the possibility of saccadic eye movements cannot be ruled out. A similar paradigm was employed by Hoff and McKeever (1979). They presented three single-digit numbers for 100 msec each and reported a left-hemisphere advantage for addition problems. However, their use of a verbal response may have introduced a bias towards left-hemisphere processing. Troup, Bradshaw, and Nettleton (1982) attempted to control for this problem by using a bimanual response. This investigation resulted in a right-hemisphere advantage (based on accuracy) for addition and multiplication problems. These studies indicate that the right hemisphere may play a role in the processing of digit stimuli, when mental arithmetic is involved. It is unclear, however, whether the right-hemisphere involvement is due to the ideographic nature of the digits or to the mechanisms involved in calculation. Two studies have employed a comparative judgment task, which involves deciding which of two digits is numerically larger. Since this task can be considered to be independent of calculation, it may provide a more accurate measure of the differential stimulus processing of digits by the two hemispheres. Besner, Grimsell, and Davis (1979) reported a lefthemisphere advantage for this task, while Katz (1980) observed a righthemisphere advantage. Katz (1980) suggested that methodological differences involving stimulus size, exposure duration and inter-trial intervals may have been responsible for the discrepancies in the findings of the two studies. Discrimination and identification tasks represent a further simplification of tasks that may be used to assess the asymmetries of digit processing. Besner, Daniels, and Slade (1982) employed random unilateral presentations of single digits and asked subjects to identify each digit using a written response. They reported a strong left-hemisphere advantage, which may have been produced by a left-hemisphere bias due to oral mediation of the written response and/or the use of only right-hand responding. Cohen’s (1975) investigation of numerical stimuli permits the most direct evaluation of Coltheart’s (1980) hypothesis since it involves a comparison of alphabetic, ideographic, and pictographic representations of the same numerical stimuli. She asked subjects to discriminate, using
250
KLEIN
AND
MC INNES
a bimanual response, between the stimuli 4 and 5, which were presented as words, digits, or dice patterns. When these stimuli were randomly presented without precues, no visual field differences were observed for any of the stimulus types. However, when subjects were provided with a cue indicating what the next stimulus type would be, a left-hemisphere advantage for words and digits (which were preceded by verbal cues) and a right-hemisphere advantage for dice (nonverbal cue) emerged. While the finding that the digit stimuli produced a left-hemisphere advantage would seem to be strong evidence against Coltheart’s (1980) hypothesis, these stimuli were cued with the word “DIGIT,” which may have introduced a left-hemisphere bias for that stimulus type (e.g., see Kershner, Thomas, & Calloway , 1977). Additionally, the use of a very limited stimulus set (only two numbers) might have encouraged a strategy of looking for distinctive features (such as the presence vs. absence of a curved line, which would be diagnostic for discriminating 4 and 5) that might favor the left hemisphere. To summarize, the results of studies that have investigated the asymmetries involved in digit processing have, up to now, been inconclusive. The methodological issues of task requirements and response selection may contribute significantly to these inconsistencies. Hemispheric biasing effects may be reduced by employing a simple task using a manual response (counterbalanced for hand) and random presentations of stimulus types to ensure that stimulus type differences are real and not due to attentional biases (see Kinsbourne, 1970). The present study was conducted to overcome some of the methodological problems associated with earlier studies of hemispheric differences in digit processing. The stimulus types (words, digits, dot patterns used on dice) employed in the Cohen study (1975) were chosen because they allow for a comparison of alphabetic, ideographic, and pictographic representations of the same numerical stimuli, thus permitting a direct evaluation of Coltheart’s (1980) right hemisphere reading hypothesis. To rule out differential attentional biases across these stimulus types, they were randomly mixed. A simple odd-even judgement task was chosen for the current study because it allows for a brief, single stimulus presentation which requires identification without the complexity of calculation. Additionally, the task permits a two-choice, unimanual response, which avoids the hemispheric biasing effects produced by a verbal response. The stimulus set used in the Cohen study (1975) was expanded to include the numbers 1 through 6, so that the use of any strategies that may have been induced by a limited stimulus set (e.g., feature extraction) could be ruled out. Based on Coltheart’s (1980) right hemisphere reading hypothesis, it was predicted that words would show a left-hemisphere advantage because they permit phonological processing. Since phonological processing is not possible for the digit and dice stimuli, it was predicted
PROCESSING NUMERICAL
FIG. I.
STIMULI
251
Examples of dot/dice, digit, and word stimuli used in the experiment.
that they would show a right-hemisphere advantage, due to the right hemisphere’s superior visuospatial processing capabilities. METHOD Subjects. Twenty-four undergraduate students (12 females; 12 males), all of whom were native English speakers, participated in the study for course credit. A handedness inventory demonstrated that all subjects were strongly right-handed. All had normal or correctedto-normal vision. Apparatus. The stimuli were displayed on a Tektronix 604 display monitor with a P31 phosphor, under the control of a PDP-II:10 computer. Subjects viewed the display from a distance of 2.5 ft. The stimuli were the numbers 1-6, displayed in the form of vertically oriented words, arabic digits, and dot patterns (those used on dice). The stimuli are shown in Fig. 1. The words were written in uppercase dot matrix letters each of which subtended about 0.17 deg horizontally and 0.2 deg vertically. The arabic numerals subtended 0.34 deg horizontally and 0.4 deg vertically. Each dot in the dice stimuli subtended about 0.08 deg, and the full matrix subtended 0.36 deg vertically and horizontally. Procedure. Each subject was required to judge whether a stimulus flashed to the right or left of a central fixation point was odd or even. The response involved pressing one of two keys with either the middle or index finger of one hand. Reaction time and accuracy were recorded for each trial. A trial proceeded as follows: a central fixation point was displayed for 2 sec. The fixation point then increased in brightness for 0.5 set, as a warning to subjects that the stimulus was about to appear. The stimulus was then presented to either the right or left of fixation for 160 msec after which the screen became blank. Subjects were allowed up to 3 set in which to make their response. Following the response, there was a 2-set intertrial interval before the next trial began. The order of presentation of stimulus type and the visual field of presentation were randomized. Each stimulus type occurred equally often in the left and right visual field.
252
KLEIN AND MC INNES -.,
I &II /
IS.91 COTS
15%
(4.9) WORos
LVF
RVF
800 2 I
t
FIG. 2. Performance as a function of visual field and stimulus type. Percentage error for each condition is shown in parentheses. Each subject received 36 practice trials following general instructions. The instructions emphasized that the subject should maintain her/his focus on the central fixation point and respond as quickly and accurately as possible to the stimulus. Following the practice trials, there were two blocks of 144 experimental trials each (with a 5-min rest period between blocks). In each block of 144 trials, each stimulus type occurred 24 times in each visual field. Each number occurred equally often in each combination of visual field and stimulus type. Subjects were required to keep their two fingers resting on the keys throughout the experiment. The response hand was counterbalanced between subjects such that half responded with their right hand and the other half with their left hand. For each of these two groups, half used their index finger for odd responses and their middle finger for even responses. The other half used the opposite arrangement.
RESULTS Mean correct reaction time (RT) and mean accuracy (% correct) were each subjected to a five-way analysis of variance (sex x hand x block x stimulus type x visual field). Analysis of the reaction time data (each entry in the ANOVA was based on up to 36 trials, depending on the number of errors) revealed main effects of block (F( 1, 20) = 13.38, p < .Ol) and stimulus type (F(2, 40) = 105.75, p < .OOOl).There were no main effects of sex, hand, or visual field. The only significant interactions involved stimulus type and visual field (F(2, 40) = 3.27, p < .OS) and sex and stimulus type (F(2, 40) = 6.11, p < .Ol). The stimulus type x visual field interaction is shown in Fig. 2. These data suggest that there is a tendency for the right visual field (RVF) to be faster with words and the left visual field (LVF) to be faster with digits and dots, exactly the pattern of results predicted by Coltheart’s hypothesis. Planned contrasts revealed that the visual field differences for each type of stimulus were not significant (words, F(1, 40) = 2.06; numerals, F(1, 40) = 3.69; dots, F(1, 40) = 1.98), but the visual field effect for the word stimuli was significantly different from that for the numerals and dots combined (F( 1,
PROCESSING
NUMERICAL
STIMULI
253
40) = 6.41, p < .025). When contrasted separately, the difference between words and numerals was significant (F(1, 40) = 5.63, p < .05) while that between words and dots was nearly significant (F(1, 40) = 4.04, p = .0512). The effect of block revealed that subjects were faster in the second block. The interaction between sex and stimulus type was due to the fact that males processed dot patterns and digits with equal speed while females were slower on the dot patterns. The analysis of accuracy revealed significant main effects of sex (F( 1, 20) = 4.78, p < .05), block (F(1, 20) = 16.13, p < .OOl>and stimulus type (F(2,40) = 14.96, p < ,001). No interactions were significant. These effects indicate that males were more accurate than females, performance accuracy improved with block, and accuracy was best for digits, worst for dot patterns, and intermediate for words. Although the interaction between visual field and stimulus type was not significant (F(1) 40) = l&3), these data are shown parenthetically in Fig. 1 to permit evaluation of the possibility that the obtained interaction in RT might be due to speed-accuracy tradeoffs. As can be seen, in each visual field comparison, the faster RTs are associated with more accurate performance, so the RT pattern is clearly nor compromised by accuracy tradeoffs. Within each type of stimulus, RTs for each numerical quantity (l-6) were combined across the two blocks and subjected to ANOVAS with visual field and quantity as variables (each entry in the ANOVA was based on up to 12 trials). Significant performance variations as a function of quantity were observed for each stimulus type: With words (F(5, 115) = 3.1, p < .025), the ordering of RTs was 6 < 1, 2, 3, 4, < 5; with numerals (F(5, 115) = 5.65, p < .Ol) the ordering was 2, 4, 6 < 1, 3, 5; and with dots (F(5, 115) = 2.99, p < .025) the ordering was 1, 4 < 2, 3, 6 < 5. However, in no case did this pattern interact with visual field (all F’s < 1). DISCUSSION
The interaction between stimulus type and visual field observed in reaction time offers support for Coltheart’s (1980) right hemisphere reading hypothesis. The word stimuli, which contain explicit phonological information, produced a RVF advantage of 12 msec. Conversely a LVF advantage of comparable magnitude was observed for the digit and dot stimuli, both of which may be characterized as lacking phonological information. These findings are open to several interpretations. If one assumes that the right hemisphere is incapable of processing linguistic information, due to its inability to perform phonological recoding, it follows that words which are presented to the LVF must be transferred to the left hemisphere in order to be processed. In this case, the 12msec visual field difference observed for word stimuli may reflect the
254
KLEIN AND MC INNES
time to transfer the graphological information from the right to the left hemisphere for phonological recoding. The corresponding decrease in accuracy (- 1.3%) may reflect the loss of information during transfer. Alternatively, the right hemisphere may utilize a strategy for identifying the words in order to make the odd/even judgment which does not require phonological recoding but which takes longer than the left hemisphere’s processing strategy. The tendency for digits and dots to be processed faster by the right hemisphere may be interpreted in a similar fashion. The observed 16msec visual field difference may reflect a transfer time of visual information from the left to right hemisphere, or it may reflect a slower and possibly different nonphonological processing strategy by the left hemisphere. These possibilities cannot be distinguished using the present data, but the fact that RT differences as a function of numerical quantity did not interact with visual field seems more consistent with an explanation based on transfer. A recently published study by Besner, Snow, and Davelaar (1986) merits discussion since it addressed the question of differential hemispheric processing of alphabetic and ideographic orthographies using a different methodology and reports different results from the present study. They compared performance on unilaterally presented single letters and digits using a verbal identification response. The major rationale seems to have been that clinical studies of split-brain patients have employed similar comparisons yielding right-hemisphere advantages for letters, but that this finding cannot be generalized to the normal population. Indeed, unlike the expectation based on the clinical literature, Besner et al. (1986) found equivalent RVF advantages for the two types of material. The authors propose a semantic distinction between single letters and numerals, with the former being void of semantic information and they argue that it is this factor which is responsible for the differential hemispheric processing in the clinical literature. While it is true that digits contain more concrete semantic information, it may be argued that single letters contain semantic associations in the form of prototype words (e.g., A = apple, C = cat). In any case, to the extent that the clinical literature has compared single letters and digits in an identification task, Besner et al. (1986) have provided a useful extension to the normal population. However, as a test of Coltheart’s hypothesis the Besner et al. (1986) study is flawed. In the first place, the appropriateness of using single letters to represent an alphabetic script must be questioned since the arbitrary assignment of pronunciation to each letter characterizes them as ideographic. A direct comparison of ideographic and alphabetic scripts is better achieved by using digits and their alphabetic counterparts (i.e., 1, one) as in the present experiment. Second, there is the very real possibility that their observed left-hemisphere advantage for letters and
PROCESSING
NUMERICAL
STIMULI
255
numerals is due to a biasing effect inherent in the use of a vocal identification response. At the very least, such a bias could override any visual field differences that may exist for the two stimulus types (see, e.g., Hellige & Sergent, 1986), and for this reason their failure to obtain a visual field by type of stimulus interaction is not particularly informative. Given our choice of materials, there is the inevitable fact that the stimuli are physically different. While the stimuli were roughly equated for visibility and were composed of the same size luminous points, the density of these points and the overall size of the stimuli were somewhat different for the word, number, and dot stimuli. In light of the proposal that the two hemispheres might differ in their ability to process high and low spatial frequencies (Sergent & Hellige, 1986) it might be suggested that the pattern we have observed is due to the relative weight of high and low spatial frequencies in the composition of our stimulus materials. We do not favor this interpretation for two reasons. First, although direct manipulation of the size of stimuli has produced changes in hemispheric asymmetries (Sergent, 1983; Pring, 1981), the spatial frequency hypothesis has not received direct support from research on the detection of sine wave gratings (Kitterle & Kaye, 1985). Second, an analysis of the spatial frequency composition of our stimuli reveals a great degree of overlap, and thus does not permit a straightforward explanation of our data in terms of the spatial frequency hypothesis. Nevertheless, although we prefer an explanation based on Coltheart’s hypothesis, we cannot rule out the possibility that physical differences between our stimuli might be partly responsible for the observed visual field by type of stimulus interaction. The role of the odd/even judgement task in mediating the observed interaction merits some discussion. It is possible that one, or the other, hemisphere is superior at performing the odd/even judgement. This could have the effect of biasing performance toward the superior hemisphere. For example, the fact that the left hemisphere was not significantly better than the right with the word stimuli suggests that the right hemisphere might be superior on this task. From the point of view of the present experiment, however, which does not permit a direct assessment of this possibility, the important observation is that the visual field effects observed depend on the type of representation of the numerical stimuli (i.e., the visual field by type of stimulus interaction). The odd/even task was chosen for the current study because it allows for a manual response, a short exposure duration, presentation of a single item, and is relatively simple. In future work it would be useful to investigate different tasks which meet these criteria, to determine the generality of the observed interaction between visual field and stimulus type in the processing of numerical stimuli.
KLEIN AND MC INNES
256
REFERENCES Besner, D., Daniels, S., & Slade, C. 1982. Ideogram reading and right hemisphere language. British Journal of Psychology, 73, 21-28. Besner, D., Grimsell, D., & Davis, R. 1979. The mind’s eye and the comparative judgement 17, 373-380. of number. Neuropsychologia, Besner, D., Snow, D., & Davelaar, E. 1986. Logographic reading: Is the right hemisphere special? Canadian Journal of Psychology, 40, 45-53. Bryden, M. P. 1982. Functional asymmetry in the intact brain. New York: Academic Press. Cohen, G. 1975. Hemisphere differences in the effects of cuing in visual recognition tasks. Journal
of Experimental
Psychology:
Human
Perception
and Performance,
1, 366-
373. Coltheart, M. 1980. Deep dyslexia: A right hemisphere hypothesis. In M. Coltheart, K. Patterson, & J. C. Marshall (Eds.), Deep dyslexia. London: Routledge & Kegan Paul. Dimond, S., & Beaumont, J. 1972. A right hemisphere basis for calculation in the human brain. Psychonomic Science, 26, 137-138. Geschwind, N. 1965. Disconnection syndromes in animals and man. Brain, 88, 327-394 and 585-644. Hatta, T. 1977.Recognition of Japanese Kanji in left and right visual field. Neuropsychologia, 15, 685-688.
Hellige, J., & Sergent, J. 1986. Role of task factors in visual-field asymmetries. Brain and Cognition,
5, 200-223.
Hoff, A. L., & McKeever, W. F. 1979. Hemispheric specialization for simple arithmetic calculation: evidence from normal subjects and from a cerebral commissurotomy patient. Clinical
Neuropsychology,
I, 24-28.
Katz, A. 1980.Cognitive arithmetic: Evidence for right hemisphere mediation in an elementary component stage. Quarterly Journal of Experimental Psychology, 32, 69-84. Kershner, J., Thomas, R., & Callaway, R. 1977. Nonverbal fixation control in young children induces a left-field advantage in digital recall. Neuropsychologia, 15, 569576. Kinsbourne, M. 1970. The cerebral basis of lateral asymmetries in attention. In A. F. Sanders (Ed.), Attention and performance III. Amsterdam: North-Holland. Kitterle, F. L., & Kaye, R. S. 1985. Hemispheric symmetry in contrast and orientation sensitivity. Perception & Psychophysics, 37, 391-396. Levy, J., & Trevarthen, C. 1977. Perceptual, semantic and phonetic aspects of elementary language processes in split brain patients. Brain, 100, 105-118. Patterson, K., & Besner, D. 1984.Is the right hemisphere literate? Cognitive Neuropsychology, 1, 315-341. Pring, T. R. 1981. The effects of stimulus size and exposure duration on visual field asymmetries. Cortex, 17, 227-240. Sasanuma, S., Itoh, M., Mori, K., & Kobayashi, Y. 1977. Tachistoscopic recognition of Kana and Kanji words. Neuropsychologia, 15, 547-553. Sergent, J. 1983. The effects of sensory limitations on hemispheric asymmetries. Canadian Journal
of Psychology,
37, 345-366.
Sergent, J., & Hellige, J. 1986. Role of input factors in visual-field asymmetries. Brain and Cognition, 5, 174-199. Smith, A. 1966. Speech and other functions after left (dominant) hemispherectomy. Journal of Neurology, Neurosurgery and Psychiatry, 29, 467-471. Troup, G. A., Bradshaw, J. L., & Nettleton, N. C. 1983. The lateralization of arithmetic and number processing: A review. International Journal of Neuroscience, 19, 321242.